Type of economic evaluations

Two economic evaluations will be conducted. The first economic evaluation will be a cost-effectiveness analysis, which will examine the primary clinical outcome measure of the trial, change in the Mood and Feelings Questionnaire (MFQ),23 where the MFQ screens for depressive symptoms. The second economic evaluation of CARIBOU-2 will be a cost-utility analysis (ie, a cost-effectiveness analysis where effectiveness is measured using a utility measure), in line with the CDA guidelines for the recommended reference case analysis,22 where the outcome measure will be QALYs. The QALY is recommended in economic evaluation studies due to its ability to be compared across different interventions and illnesses/disorders.22 Both economic evaluations will be undertaken at two time points, 24 and 52 weeks post-enrolment follow-up, using the sample with non-missing data (where adolescents who are lost to follow-up over the course of the trial or with missing data on outcomes and/or costs will be excluded) as well as the sample with imputed data.

Study population

Participant recruitment (planned n=300) will occur over 4.5 years, from February 2022 to September 2027, at four to six sites (hospitals and community-based mental health agencies) across southern Ontario and Alberta, where youth often receive outpatient mental healthcare. Adolescents will self-refer or be referred by a third party (eg, doctors, school counsellors, caregivers) to the site, and then recruited after their intake. Site staff (eg, intake workers, clinicians) will assess the youth, including the use of the MFQ23 to screen for depression and inclusion/exclusion criteria. Informed consent will be obtained from all study participants by a study research assistant.

The trial will include adolescents between the ages of 13 and 18 years, inclusive, who express that ‘depression’ (or some synonym of depression) is a primary concern, where clinician or intake staff agrees that depressive symptoms are a primary treatment target, who have an MFQ score ≥22 at two sequential visits (screening and baseline assessment), who are either a new referral to the clinic in the past 3 months or, if previously received treatment at the clinic, had a period of 3 months without treatment in the past 6 months, and who are able to speak and read English. The trial will exclude youth with known or highly suspected presentations of psychotic symptoms (eg, hallucinations) that are persistent, affect functioning and have observable effects on behaviour; those with severe substance use disorder, bipolar disorder, intellectual disability, severe eating disorder and imminent risk of suicide requiring hospitalisation as per judgement of the assessing clinician; and those unable to provide informed consent to the study for any reason.

If the adolescent agrees, caregivers will also be asked to participate in the study. Other than fluency in English and capacity to make decisions regarding consenting to research, there are no other inclusion or exclusion criteria for caregiver participation.

TAU will be provided in the same hospital/community mental health agency and may or may not include referral to psychotherapy and/or parental support; psychiatric care and the use of psychotropic medication is permitted.

Perspective

We will adopt the perspective of the publicly funded healthcare payer (ie, the Ontario Ministries of Health and Long-Term Care and the Alberta Ministry of Health), in line with the CDA guidelines for the recommended reference case analysis,22 which includes all health system costs. According to the CDA guidelines, when a broader societal perspective is of interest to the decision-maker, the impact of the intervention on time lost from paid and unpaid work by both patients and informal caregivers due to illness, treatment, disability or premature death should be included in an additional non-reference case analysis.22 Therefore, we will also undertake an additional analysis (ie, a non-reference case analysis), where we will adopt a modified societal perspective, and caregiver time costs and lost income due to appointments will be considered, in addition to health system costs. The results of the non-reference case analysis will be reported separately from the reference case analysis for each outcome, in line with the CDA guidelines.22 The inclusion of caregiver time costs and lost income due to appointments in the non-reference case analysis will shed light on the impact of the intervention beyond the healthcare system (ie, the reference case) as well as its impact on caregivers.

Time horizon and discounting

The time horizon of the analysis will be 52 weeks post-enrolment, the length of participant involvement in the trial. This time horizon allows time for each component of the intervention to be completed if indicated, while accounting for wait times. When the time horizon is less than 1 year, discounting is not needed.22

Intervention resource use and costs

We will record all resources used by patients during the delivery of the intervention; these will include the time of personnel involved in the assessment of patients, delivery of education sessions, and psychotherapy, the number of medications (eg, fluoxetine, sertraline, escitalopram and duloxetine) delivered by personnel, the number of supplies and services, training of staff delivering the intervention and programme resources (eg, educational materials) related to the intervention. Subsequently, we will employ a micro-costing approach27 to estimate the total costs associated with the delivery of the intervention (ie, we will monetise the intervention-related resource use). We will obtain the relevant unit costs for each resource from the Ontario Health Insurance Plan and the Alberta Schedule of Medical Benefits fee schedules, community mental health agencies, hospital records (to estimate the salary of professionals involved and the supplies and services) and pharmacy records, the Ontario Drug Benefit Formulary and the Alberta Pharmaceutical Information Network (to estimate the cost of medications).

Resource use

Data on health system-related resource use post-intervention for Ontario will be obtained through ICES (formerly known as the Institute for Clinical Evaluative Sciences), an independent non-profit research institute in Toronto, Ontario, which holds health records for all health services covered under the Ontario public healthcare system (eg, physician visits, emergency department visits, acute care hospitalisations, psychiatric hospitalisations). We will use a custom health service utilisation tool, developed by the research team and based on an existing tool,28 to measure all health system-related resource use post-intervention for Alberta. This health service use tool will also be used to obtain post-intervention data on time spent to obtain care for both youths and caregivers, where applicable, for both provinces, and lost time away from work to obtain care for both youths and caregivers, where applicable, for both provinces. Trained research analysts will administer the health service utilisation tool to adolescents and caregivers at baseline and 12, 24, 36 and 52 weeks. Our data collection methods will also collect information on significant adverse events, such as psychiatric hospitalisations, episodes of self-harm with potential for high lethality and completed deaths by suicide, which will be reported if/when these instances occur.

Cost estimation

To estimate total costs, we will apply patient-level costing to value all resource use post-intervention for each adolescent (ie, direct health system costs, direct out-of-pocket costs), where the respective number of units reported (eg, number of visits, number of medications consumed) described beforehand will be multiplied by the respective unit cost.29 The same approach will be applied to estimate time costs (eg, time spent to obtain care) and indirect costs (eg, lost work income due to appointments) for both youths and caregivers. The unit costs will be obtained from the Canadian Institute for Health Information, the Ontario Health Insurance Plan fee schedule, the Alberta Schedule of Medical Benefits and Statistics Canada, among other sources. All costs will be expressed in 2027 Canadian dollars using Statistics Canada’s Consumer Price Index for Health and Personal Care.30

Sensitivity analyses

Several sensitivity analyses will be undertaken to account for potential biases. A systematic review found that utility values reported for adolescent depression ranged from 0.495 to 0.81.25 The main analysis will use a utility value of 0.81 for adolescents with depression; however, a deterministic one-way sensitivity analysis with a utility value of 0.495 will be undertaken to test the robustness of the results of the cost-utility analysis. As described beforehand, we will exclude participants with missing data from the main analysis; however, we will examine the sociodemographic and clinical characteristics of adolescents included in the analyses and those in the full sample to assess the impact of excluding those with missing data. In addition, we will rerun the cost-effectiveness and cost-utility analyses with imputed data on outcomes and costs using multiple imputation by chained equations.35 We will conduct deterministic one-way sensitivity analyses to determine the robustness of our results to changes in intervention and health service unit costs in instances where precise unit costs cannot be obtained; in these cases, we will use 95% CIs to determine the range (ie, high and low-cost scenarios) to be used in the sensitivity analyses. In addition, we will compare the estimates from the multivariable generalised linear mixed model to the unadjusted mean values as well as the estimates obtained from an ordinary least squares model. We will use pattern-mixture models36 to understand how potential outliers, and their exclusion, affect our findings as well as any deviations from distributional assumptions and the impact of baseline variables.

Uncertainty

We will estimate the multivariable generalised linear mixed models with non-parametric bootstrapping (namely 1000 bootstrap replications) to produce standard errors and p values for each incremental cost and effect, while adjusting for sampling uncertainty. We will examine uncertainty using cost-effectiveness planes and cost-effectiveness acceptability curves (CEACs), in line with a net benefit framework.37 Cost-effectiveness planes depict the uncertainty regarding the cost and effect estimates; this is done by plotting the respective estimated bootstrapped values.32 CEACs provide an alternative to the ICER CIs; they are obtained from the joint distribution of incremental costs and effects from the non-parametric bootstrapping of the observed data. The CEAC shows the probability that a given intervention is cost-effective compared with the comparator for several different values that a decision-maker is hypothetically willing to pay for a unit improvement in a given health outcome.38 39 We will calculate a series of net benefits for each individual for a range of willingness to pay values for a QALY and then compare these to $50 000 CAD, which is the cost-effectiveness threshold commonly used for decision-making in Canada.40 We will obtain the coefficients of the differences in the net benefits between the intervention and TAU groups through bootstrapped linear regressions, which will control for the variables included in the main analysis (eg, demographics and baseline clinical measures) as well as the baseline variables of interest, such as gender, age, ethnicity and race, to account for any potential differences between the intervention and TAU groups at recruitment. We will then examine these coefficients to determine the proportion of instances in which the net benefit of the intervention group is greater than that of the TAU group for each willingness to pay value.41 Subsequently, we will plot these proportions to obtain CEACs for each cost-effect combination. All data analyses will be undertaken using Stata V.12.